Automated collaborative filtering system

• US 5,790,426 A
• Filed: 04/30/1997
• Issued: 08/04/1998
• Est. Priority Date: 04/30/1996
• Status: Expired due to Fees

• Current Assignee: Virtual Development Corporation
• Original Assignee: Athenium, LLC (Athenium Analytics)
• Inventors: Robinson, Gary B.
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): Bui, Bryan
• Application Number:
  US08/848,317
• Time in Patent Office: 461 Days
• Field of Search: 705/27-29, 705/10, 705/6, 707/1, 707/3, 364/253.1, 364/282.1, 364/286, 364/554, 364/DIG. 1, 364/553, 364/555, 273/161, 395/187.01, 395/188.01
• US Class Current: 702/179
• CPC Class Codes: G06F 16/335 Filtering based on additional data, e.g. user or group profiles filtering in web context G06F16/9535, G06F16/9536
• CPC Class Codes: G06Q 10/025 Coordination of plural reservations, e.g. plural trip segments, transportation combined with accommodation
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

An automated collaborative filtering (ACF) system for recommending at least one item to a first user based on similarity in preference of the user as compared with other users. The ACF system stores rating data for items provided by users of the system. Upon request of the first user, the system determines similarity values for the first user as compared with other users that have provided rating data for items that the first user has also rated. Based on the similarity values, a subgroup of users is selected that is then used to provide recommendations to the first user.

Claims

1. A method of providing a recommendation of at least one item in a group of items to a first user in a group of users based on ratings of the item and other items in the group of items by other users in the group of users and ratings of at least some of the other items by the first user, comprising the steps of:

providing rating data for items in the group of items from at least some of the users in the group of users, wherein each item has a rating distribution based on the ratings by the at least some of the users in the group of users;

determining at least one item that the first user and a second user have provided ratings for;

determining transformed ratings for the at least one item based on the rating distribution of the at least one item;

determining a median transformed rating from rating provided by the first user for the at least one item based on the rating distribution;

determining a range of a randomized transformed rating for the rating of the at least one item provided by the second user;

determining a probability that a rating for the at least one item provided by a randomly chosen third user is at least as statistically close to the median transformed rating of the first user as the randomized transformed rating for the second user;

determining a test statistic for the first and second user from the probability; and

determining a similarity value from the test statistic.

2. The method according to claim 1, wherein the similarity is calculated using a Z-score.

3. The method according to claim 1, wherein the similarity value is calculated using a chi-square distribution.

4. The method according to claim 1, wherein the test statistic is weighted.

5. The method according to claim 4, wherein the test statistic is weighted based on the similarity of items.